-
Ambition: a Europe-wide benchmark -----------------------------------------------
The team led by Cédric Villani spared no effort in preparing its report, available at https://www.aiforhumanity.fr : hearings with hundreds of experts, consideration of numerous contributions from the public, and great care devoted to the quality and clarity of the prose... Precise in its terminology and clear in its account of the current situation, this lengthy document includes five sector-specific "focus" sections: education, health, agriculture, transport, defence and security. Its authors tell us that the report was "written to advise the government, with action and effectiveness firmly in view", with the ambition of becoming a benchmark across Europe.
The report also proposes tripling the number of people trained in AI in France by 2020, which will require new courses to be created. Finally, it recommends pursuing an ambitious policy to bring more women into the AI sector, raising their share of the field to 40% by 2020.
-
The role of data -------------------
Around forty pages are devoted to the raw material of artificial intelligence: data. The report recommends making data accessible to everyone and encouraging its circulation and sharing through dedicated platforms developed in collaboration with other European countries. It also recommends establishing a world-class research network, with special resources enabling it to counter the brain drain currently observed towards both major foreign organizations and private-sector companies. Finally, the report does not overlook the ethical questions raised by the profound changes that artificial intelligence may bring about in our society.
-
How much will it cost? ----------------------
Deliberately geared towards action and effectiveness, the Villani report makes a host of concrete proposals. Nevertheless, questions remain. The first concerns funding: for computing resources and salaries, for the human and social costs arising from changes in working conditions, and for changes to education. While the 2017 "France IA" report (available online) included cost estimates, this one does not: it repeatedly acknowledges that funding may prove problematic, but never provides further details. The report also seems to favour the upstream side, namely research, over the downstream side—the effects on society. One proposal for establishing AI in France is to give everyone a general grounding in algorithms from the first year of primary school. As for data, a subject covered extensively, the proposal to broaden access to it could conflict with the recently introduced General Data Protection Regulation. How will these two principles be reconciled?